Preferential oxidation of carbon monoxide in the presence of hydrogen (PROX) over ceria-zirconia and alumina-supported Pt catalysts

Article abstract of DOI:10.1016/j.jcat.2004.04.017

A crucial requirement for the operation of proton exchange membrane fuel cells is to feed CO-free hydrogen to the anode. This can be achieved using catalysts that can selectively oxidize CO in the presence of excess hydrogen. Preferential CO oxidation (PROX) in the presence of hydrogen over Pt/Ce_xZr_x-1O₂ (x = 0, 0.15, 0.5, 0.68, 1) catalysts was investigated; results were also compared with those observed on Pt/Al₂O₃ catalysts. Ceria-supported Pt catalysts were more active than Pt/Al₂O₃ in both CO and H₂ oxidation. The result was a sharp light off at 90°C for oxygen conversion. Chloride ion-containing Pt/CeO₂ catalysts showed lower performance in the PROX reaction, especially at low temperatures. Four types of reaction mechanisms were suggested for Pt/Ce_xZr_x-1O₂ samples: competitive Langmuir-Hinshelwood CO and hydrogen oxidation on Pt particles; noncompetitive Langmuir-Hinshelwood mechanism on the metal/oxide surface; hydrogen oxidation on the support; and water-gas-shift reaction at high temperatures. The second reaction route predominated at (90°-130°C, and was preferential to CO oxidation rather than hydrogen-oxygen reaction.

Full text of DOI:10.1016/j.jcat.2004.04.017

Journal of Catalysis 225 (2004) 259–266
www.elsevier.com/locate/jcat
Preferential oxidation of carbon monoxide in the presence of hydrogen
PROX) over ceria–zirconia and alumina-supported Pt catalysts
(
a,b,∗
a,c
a
Attila Wootsch,
Claude Descorme, and Daniel Duprez
a
LACCO-UMR 6503, CNRS-Université de Poitiers, 40, Avenue du Recteur Pineau, F-86022 Poitiers cedex, France
b
Institute of Isotope and Surface Chemistry, CRC, HAS, Budapest PO Box 77, H-1525, Hungary
c
Institut de Recherches sur la Catalyse (IRC), CNRS, 2, Avenue Albert Einstein, F-69626 Villeurbanne cedex, France
Received 11 December 2003; revised 16 March 2004; accepted 15 April 2004
Abstract
One crucial requirement for the operation of proton exchange membrane fuel cells (PEMFC) is to feed carbon monoxide free hydrogen to
the anode. This need can be achieved by using catalysts able to selectively oxidize CO in the presence of excess hydrogen. Herein we report
the preferential CO oxidation (PROX) in the presence of hydrogen over Pt/CexZr
O (x = 0, 0.15, 0.5, 0.68, 1) catalysts. A comparison
x−1
2
◦
with results observed on a Pt/Al O catalyst is also presented. We examined the effect of temperature (90–300 C) and O excess (λ = 0.8–2).
2
3
2
Ceria-supported platinum catalysts were more active than Pt/Al O in both CO and H oxidation. The result was a sharp “light off” around
2
3
2
◦
9
0 C for the oxygen conversion. The maxima, which appeared in the CO conversion and in the selectivity toward CO oxidation as a function
of temperature on Pt/Al O , did not show up in the case of ceria-supported samples. Chloride ion-containing Pt/CeO catalysts showed lower
2
3
2
performances in the PROX reaction, especially at low temperatures. Four types of reaction mechanisms were suggested for Pt/CexZr
O
2
x−1
samples: (i) competitive Langmuir–Hinshelwood CO and hydrogen oxidation on Pt particles, (ii) noncompetitive Langmuir–Hinshelwood
mechanism on the metal/oxide interface, (iii) hydrogen oxidation on the support, and (iv) water–gas-shift reaction at high temperatures.
◦
The second reaction route predominated at low temperatures (90–130 C) and was found preferential to CO oxidation rather than hydrogen–
oxygen reaction. This process was selectively blocked by Cl ions. The possible application of Pt/CeO and Pt/Al O catalysts was discussed.
2
2 3
Ceria appeared as a suitable support for the preferential CO oxidation catalysts at low temperatures.
2004 Elsevier Inc. All rights reserved.

Keywords: Preferential CO oxidation; PROX; Platinum; Pt/CeO ; Ceria; Ceria–zirconia; Oxygen mobility; Hydrogen purification; Fuel cell
2
1
. Introduction
or partial oxidation (POX) of natural gas, light oil fractions,
and alcohols [5–7]. Unfortunately, a noticeable amount of
Application of fuel cell-powered systems to transporta-
2 2
CO, ca. 5–15%, is formed together with H , H O, and
tion recently has received increasing attention because of
their theoretical high fuel efficiency and low environmen-
tal impact [1,2]. Numerous types of fuel cells have already
been developed [1,2]. Among them, one of the most promis-
ing is the proton exchange membrane fuel cell (PEMFC)
fueled with hydrogen [1–3]. However, depending on the type
of anode, the CO concentration in the hydrogen feed must
be under 1–100 ppm [1,4]. Hydrogen is usually produced
by steam reforming (STR), autothermal reforming (ATR),
CO . A subsequent water–gas-shift (WGS) stage reduces the
amount of CO to 0.5–1% [4–6]. This high amount of CO
can be removed by preferential oxidation (PROX) and/or by
methanation of CO using mainly Pt and Pt–Ru bimetallic
catalysts [8,9].
Oh and Sinkevitch published the first general work on
the PROX in 1993 [10]. They tested a variety of noble
metals, such as Ru, Rh, Pt, and Pd on alumina support as
well as other possible oxidation catalysts, such as Co/Cu,
Ni/Co/Fe, Ag, Cr, Fe, and Mn, respectively. Platinum,
rhodium, and ruthenium were found to be the most appro-
priate for the PROX reaction: CO was totally converted to
CO2 but a noticeable amount of hydrogen was also simul-
2
*
Corresponding author. Tel.: +36-1-392-2222/3172; fax: +36-1-392-
2533.
E-mail address: wootsch@alpha0.iki.kfki.hu (A. Wootsch).
0021-9517/$ – see front matter  2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2004.04.017

260
A. Wootsch et al. / Journal of Catalysis 225 (2004) 259–266
taneously consumed in the presence of excess oxygen. An
(i) The “operating window” is too narrow. Platinum cat-
alysts can be used efficiently in a small temperature
◦
“
optimum” reaction temperature was 100 C for Rh and
◦
◦
Ru [10,11] and around 170 C for Pt [10,12]. Co/Cu/Al2O3
range (170–200 C) and in the presence of excess oxy-
and Ni/Cu/Al2O3 were active in the reaction only above
gen, λ = 2.2 [12].
◦
2
50 C, with a 40–50% selectivity at λ = 2.
(ii) The selectivity toward CO oxidation must be increased
to minimize the hydrogen lost.
(
Several catalytic systems have been tentatively tested in
iii) The effect of the fluctuations in the inlet gas concentra-
tion must be compensated.
the PROX reaction so far. The most extensively studied
metal is still platinum [12–21], but gold was also found to be
a suitable catalyst for the reaction [22–27]. Certain studies
deal with the comparison of these two metals [28,29]. Gold
catalysts are usually less active but more selective than the Pt
ones. Unfortunately, gold also deactivates more rapidly upon
long-term operation [25,29].
In our understanding, the CO oxidation in reductive at-
mosphere could be analogous to three-way catalysis (TWC)
in the case of automotive exhaust posttreatment. In that
case cerium–zirconium mixed oxides are added to the sup-
port to promote the oxidation reactions under oxygen-
poor conditions [34–36]. Ceria supported platinum [37,38],
rhodium [39,40] as well as palladium [41] catalysts are also
remarkably active in the low-temperature oxidation of CO.
Ceria–zirconia-supported noble metal catalysts, prepared
and characterized in our laboratory, were able to oxidize
carbon monoxide even in the absence of oxygen [42,43].
Herein we report preferential CO oxidation in the pres-
ence of excess hydrogen over Pt/CexZrx−1O2 (x = 0, 0.15,
Other formulations were also tested such as Pt–Sn al-
loy [30], Ru catalysts [31,32], and Pt, Ru, and Rh containing
γ -Al2O3–SiO2 microporous membranes [33]. Ruthenium
was found to be superior to platinum. It was active at a lower
temperature and resulted in higher selectivities toward CO2
formation. Nevertheless, methane formation was observed
◦
above 250 C over Ru/Al2O3 and hydrogen was again con-
sumed [31].
The usual reactor setup, at a laboratory scale, is a flow re-
actor operated at atmospheric pressure [10–33]. A two-stage
reactor was also reported where oxygen was added to the re-
action mixture before and in between the two stages [14].
Half-industrial PROX reactors were developed using Pt–Ru
bimetallic catalysts [8,9]. In these cases the chemical process
was not described in detail, but considered as mainly oxi-
dation accompanied by methanation and a water–gas-shift
reaction.
0
.5, 0.68, 1) catalysts in comparison with results observed
on Pt/Al2O3. Our goal was to study the effect of the reac-
tion conditions (T , excess O2, λ), to compare the different
catalysts and to get information on the reaction mechanism.
2
. Methods
2
.1. Catalysts
Existing PROX catalysts might lower the CO concen-
tration in a hydrogen-rich stream under restricted reaction
conditions [8,9]. In fact, some problems hamper their appli-
cation:
An overview of the catalysts is presented in Table 1.
Pt/Al O was prepared by impregnation of H PtCl on a
2
3
2
6
2
γ -Al2O3 support (GFS, BET = 210 m /g, precalcined at
Table 1
Characteristics of the Pt catalysts tested in the PROX reaction
a
b
c
d
e
Catalyst
y (%)
D (%)
m (mg)
BET
2
Cl content
(%)
−1
2
(
m g
)
Cl/Pt
(atom/nm )
Prepared from H PtCl
Pt/Al O
2 3
0.6
1
75
34
112
102
207
12
0.02
0
0.2
–
0.02
–
2
6
Prepared from Pt(NH ) (OH)
Pt/ZrO
3
4
2
2
Cl-free ceria-containing
samples, prepared from
Pt/CeO
1
1
1
1
53
67
50
48
100
128
103
116
28
41
40
33
0
0
0
0
–
–
–
–
–
–
–
–
2
Pt/Ce
Pt/Ce
Pt/Ce
Zr
Zr
O
O
O
0
.68 0.32
2
2
2
Pt(NH ) (OH)
3
4
2
0.50 0.50
Zr
0.15 0.85
f
f
f
Impregnation of Cl ions
Pt/CeO (Cl/Pt = 1)
1
1
1
52
53
51
112
103
98
28
27
28
0.15
0.29
0.56
0.8
1.6
3.1
0.9
1.8
3.4
2
Pt/CeO (Cl/Pt = 2)
2
Pt/CeO (Cl/Pt = 4)
2
Prepared from H PtCl
2
Pt/CeO
1
54
117
26
0.66
3.6
4.3
6
2
a
y, metal content of the catalyst, m/m%.
D, metal dispersion of the catalyst.
m, mass of catalyst placed in the reactor.
b
c
d
e
Measured after impregnation.
2
Cl content was measured by elemental analysis (Service Central d’Analyses du CNRS, France). The value in atom/nm is calculated from Cl content
(
m/m%) and BET surface area.
f
Cl-free Pt/CeO sample was impregnated by Cl ions in amounts corresponding nominally to 1, 2, and 4 Pt-atom eq.
2

A. Wootsch et al. / Journal of Catalysis 225 (2004) 259–266
261
◦
4
50 C, DAIR = 30 mL/min). The impregnated sample was
2.3. Calculations
◦
washed by bidistilled water, dried overnight at 120 C, cal-
◦
cined at 300 C for 4 h in flowing air (DAIR = 30 mL/min),
Two competing reactions were taken into consideration:
the CO oxidation (2CO + O2 → 2CO2) and the hydrogen
oxidation (2H2 + O2 → 2H2O). The inlet amount of the
◦
and finally reduced at 500 C for 4 h in flowing H2 (DH =
2
3
0 mL/min). Dispersion was determined by chemisorption
in
in
in
in
of H2 at room temperature.
different gasses is known: n , n , n , n . The follow-
CO O2
H2 He
The preparation and the characterization of the ceria–
zirconia supports and the corresponding catalysts have al-
ready been described elsewhere [34,42–44]. Briefly, the sup-
ports (Rhodia Catalysts & Electronics) were precalcined at
ing equations can be established from the stoichiometry and
mass balances:
H,
in
out
out
◦
n
= nH2 + n
,
(1)
(2)
(3)
(4)
9
00 C for 6 h (DAIR = 30 mL/min) and impregnated using
H2
H2O
a Cl-free metal salt, Pt(NH3)4(OH)2. The received samples
were dried at 120 C for 24 h and calcined for 4 h at 500 C
C,
◦
◦
in
CO
out
out
CO
◦
n
= n + n
,
in flowing air (DAIR = 30 mL/min) and reduced at 400 C
CO
2
for 4 h in flowing H2 (DH = 30 mL/min). Finally, the sam-
2
O,
ples were characterized by XRD, TEM, BET, and elemental
◦
in
O2
in
CO
out
out
CO2
out
out
analysis as well as low temperature (−85 C) H2 adsorp-
2 · n + n = 2 · nO2 + 2 · n
He,
+ n + n
,
CO
H2O
tion [45] to determine the metal dispersion [42–44].
In order to study the effect of Cl ions on the performance
of Pt/CeO2 catalysts in the PROX reaction, a Pt-on-ceria
catalyst has been prepared from H2PtCl6 and pretreated as
above. Other Cl-containing samples were prepared by im-
pregnation of the Cl-free Pt/CeO2 catalyst by chloride ions,
using HCl. Charges of 1 g of this catalyst were immersed in
solutions of 0.1 M HCl. The Cl content of the solutions was
set at 1, 2, and 4 Pt-atom eq. After evaporation of the solu-
in
He
out
n
= nHe
.
out
H2
out
CO
out
O2
out
,
There are 6 different outlet flows n , n , n , n
He
out
2
out
CO
n
, and n , respectively. Because of the 4 linearly inde-
H O
2
pendent equations (1)–(4), measurement of 2 linearly inde-
pendent outlet flow rates could define all the outlet parame-
ters. In practice, we determined the outlet amount of CO and
◦
2 2
O and we used the CO concentration to check the mass
tion, the cake was dried at 120 C overnight and calcined at
◦
balance. Analysis of hydrogen and water only gave approx-
imate values so that the above-noted equations were used to
determine their exact concentrations.
The total conversion was defined as the oxygen consump-
tion
4
00 C in flowing air for 4 h (DAIR = 30 mL/min) and re-
◦
duced at 400 C for 4 h in flowing H2 (DH = 30 mL/min).
2
The efficiency of chlorination was between 60 and 80% (Ta-
ble 1). Obviously, some of the Cl left the surface during the
pretreating procedure. The chlorination did not really affect
the BET surface of the samples (Table 1).
in
out
O2
n
− n
O2
O2
Xtotal = XO =
· 100 (in %).
(5)
2
in
n
2
.2. Reactor setup
The selectivity was calculated as the ratio of the desired
Catalytic tests were carried out in an atmospheric con-
reaction (CO oxidation) to the overall reactions, H and CO
2
tinuous flow reactor system. Tubing and connections were
made from stainless steel. Analytical grade (Alphagas-1)
cylinders of hydrogen, helium, and CO/He 1/1 as well
as O2/He 1/1 mixtures were used as inlet gases and con-
trolled by Brooks mass-flow controllers previously cali-
brated. Product analysis was preformed by two gas chro-
matographs:
consumption. Such a definition of the selectivity is also valid
when methane formation occurs
ꢀ
CO
S = ꢀ
· 100
CO + ꢀH2
in
CO
out
out
n
− n
CO
in
=
· 100 (in %).
(6)
in
out
n
− n + n − nH2
CO
CO
H2
The selectivity can also be defined as the ratio of the oxy-
gen transformed into CO2 to the total oxygen consumed.
When no methane is formed: Eq. (6) = Eq. (7). Thus, from
Eqs. (1)–(3) and (6):
(
i) one equipped with a polar column, Poropak Q, to sepa-
rate CO2 and H2O from the other outlet gases, and
(
ii) the other equipped with a 5 Å molecular sieve filled col-
umn, for CO and O2 separation.
out
n
CO2
S =
· 100 (in %).
)
(7)
In both cases TCD detectors were used. Helium was used
as carrier gas for the GCs. Principally, CO2 and H2O were
detected as the only products. Methane formation did not ap-
pear in our experimental conditions. The total gas inlet was
100 N mL/min, containing 70% H2, 5% CO, 2–5% O2 (oxy-
gen excess, λ, from 0.8 to 2), and 17–20% He as a balance.
in
out
2
· (n − nO2
O2
The CO conversion, in fact, can be defined as
in
out
CO
out
CO
n
− n
CO
n
CO
2
XCO =
· 100 =
· 100 (in %).
(8)
in
in
n
n
CO

262
A. Wootsch et al. / Journal of Catalysis 225 (2004) 259–266
Lambda (λ) is the oxygen excess factor. By definition
3. Results
Fig. 1 presents the results obtained over Pt/Al2O3. These
in
in
in
2
· n
2 · c
2
· pO
pCO
2 · [O ]
O2
2
2
O2
results corresponded well to typical literature values [10,12,
16]. The oxygen conversion increased slightly but signifi-
cantly with increasing λ, probably due to the positive oxygen
order earlier observed for Pt catalysts [12]. Nevertheless,
platinum showed a natural selectivity toward CO oxidation
as lower λ values resulted in higher selectivity. The selectiv-
ity showed a maximum as a function of temperature at all λ
values (Fig. 1).
λ =
=
=
=
c
CO
.
(9)
in
in
out
n
[CO]
CO
λ = 1 means that oxygen is present in stoichiometric
amount, λ > 1 corresponds to an oxygen excess compared
to pure CO oxidation, and λ < 1 corresponds to a deficit in
oxygen compared to pure CO oxidation.
It must be noted that the product of the total conversion
Platinum on ceria catalysts showed a different behavior
(
XO ) with the selectivity (S) does not directly result in the
2
(
Fig. 2). The oxygen conversion was very high, even at low
CO conversion. The oxygen excess must be taken into ac-
count. Then, from Eqs. (5), (7)–(9):
◦
temperatures. We observed a sharp “light off” around 90 C
(
Fig. 2), according to the observations in the CO oxidation
reaction [37–39]. The maximum in the selectivity as a func-
tion of temperature was observed at a much lower temper-
ature than on Pt/Al O or did not even appear (cf. Figs. 1
2 3
S · XO
2
XCO =
· λ (in %).
(10)
1
00
Fig. 1. Preferential oxidation of CO on the Pt/Al O catalyst. Oxygen conversion (1), selectivity toward CO oxidation (E), and CO conversion (") as a
2
3
function of temperature at different oxygen excess, (a) λ = 0.8, (b) λ = 1, (c) λ = 1.5, (d) λ = 2.
Fig. 2. Preferential oxidation of CO on the Cl-free Pt/CeO catalyst. Oxygen conversion (1), selectivity toward CO oxidation (E), and CO conversion (") as
2
a function of temperature at different oxygen excess, (a) λ = 0.8, (b) λ = 1, (c) λ = 1.5, (d) λ = 2.

A. Wootsch et al. / Journal of Catalysis 225 (2004) 259–266
263
Table 2
PROX reaction on the different Pt catalysts at selected temperatures at λ = 1
and λ = 2 X , CO conversion; X , oxygen conversion; S, selectivity
CO
O
2
toward CO oxidation
Catalyst
λ = 1
CO
λ = 2
CO
X
X
O
S
X
X
O
2
S
2
(%)
(%)
(%)
(%)
(%)
(%)
◦
(
a) T = 100 C
Pt/Al O
0.7
1.6
98
93
99
98
43
80
79
70
58
60
10
95
59
76
60
98
12
98
98
97
99
98
40
48
30
39
30
50
2
3
Pt/CeO (Cl-free)
78
74
69
57
58
2
Pt/Ce
Pt/Ce
Pt/Ce
Zr
O
O
O
0
0
0
.68 0.32
2
2
2
Zr
.50 0.50
Zr
.15 0.85
Pt/ZrO
95
2
◦
(
b) T = 150 C
Pt/Al O
33
61
55
44
38
29
60
99
93
99
98
97
55
62
59
45
39
30
98
65
34
53
36
88
95
99
97
98
99
98
51
33
18
27
18
45
2
3
Pt/CeO (Cl-free)
2
Pt/Ce
Pt/Ce
Pt/Ce
Zr
O
O
O
0
0
0
.68 0.32
2
2
2
Zr
.50 0.50
Zr
.15 0.85
Pt/ZrO
2
◦
(
c) T = 200 C
Pt/Al O
68
44
43
33
31
23
98
97
94
98
99
98
71
45
46
34
31
24
61
55
32
37
34
75
97
99
96
97
98
98
32
28
17
19
17
38
2
3
Pt/CeO (Cl-free)
Fig. 3. Effect of chlorine on the catalytic performances of 1% Pt/CeO cata-
2
lysts in the PROX reaction at 100 C (E) and 200 C (2), at oxygen excess
2
◦
◦
Pt/Ce
Pt/Ce
Pt/Ce
Zr
O
O
O
0
0
0
.68 0.32
2
2
2
Zr
λ = 1 (left) and λ = 2 (right). The Pt/CeO catalyst at 0 point was prepared
.50 0.50
2
Zr
.15 0.85
from Pt(NH ) (OH) precursor; this sample was impregnated by chloride
3 4 2
Pt/ZrO
ions, using HCl in amounts corresponding nominally to 1, 2, and 4 Pt-atom
eq. Results are also shown on a Pt/CeO sample prepared from H PtCl
2
2
2
6
precursor. The x axis represents the measured Cl content of the catalysts
Table 1).
(
and 2). Accordingly this catalyst adequately converted CO
at temperatures around 100 C.
◦
Different cerium–zirconium mixed oxides as well as
zirconia-supported Pt catalysts have also been examined in
the PROX reaction (Table 2). It must be considered that
zirconia is not an active support for the oxygen storage as
opposed to ceria or cerium–zirconium mixed oxides [36].
However, ZrO2 has a better oxygen mobility than alumina,
even if it is much less than CeO2–ZrO2 [46]. Moreover,
some oxygen vacancies could exist on zirconia at the pe-
riphery of Pt particles. Consequently, the behavior of ZrO2
should be intermediary between that of alumina and those of
droxyl groups, which were claimed to be one key parame-
ter responsible for the possible migration of chemisorbed
oxygen species [39,48]. Fig. 3 shows the performance of
a Pt–ceria system in the PROX reaction as a function of
Cl content on the samples. Oxygen conversion was hardly
affected by the presence of Cl (Fig. 3). The selectivity to-
ward carbon dioxide formation, in turn, diminished over
chlorinated samples. The effect of chloride ions, as a rule,
was more important at low temperatures and at low oxy-
gen excess (Fig. 3). The CO conversion sharply decreased
in the case of the catalyst prepared from H2PtCl6 precur-
sor (Fig. 3). Obviously, the Cl concentration in the vicinity
of Pt particles is higher when a Cl-containing precursor is
used, while random Cl deposition on the support can be
suggested when chlorine is impregnated on the catalyst af-
terward. These results also point to the important role of
oxygen mobility that will be discussed in detail in the next
section.
CexZrx−1O2 (x ꢀ= 0) supports (Table 2). Pt on CexZrx−1O2
catalysts showed the same kind of behavior as Pt/CeO2.
Table 2 illustrates that the oxygen conversion over these
catalysts, active in the oxygen storage, was always around
1
00%. These catalysts could convert remarkable amount of
carbon monoxide at low temperature, even at low oxygen
excess (Table 2). At higher temperature, however, they were
too active to be able to selectively oxidize CO. In this range
Pt/Al2O3 resulted in higher selectivity toward CO oxidation.
Both Pt/Al2O3 and Pt/ZrO2 catalysts showed a maximum
selectivity as a function of temperature, but this optimum
appeared at lower temperature in the case of the ZrO2 sup-
port.
4. Discussion
4.1. Mechanistic consideration
The presence of chloride ions dramatically decreases
the oxygen spillover and back-spillover rates by forming
CeOCl species on the ceria support [47]. Such CeOCl
species formation decreases the number of available hy-
The overall picture is different, as a rule, on the supports
that are active or inactive in the oxygen storage. It might
derive from a different reaction mechanism.

264
A. Wootsch et al. / Journal of Catalysis 225 (2004) 259–266
The maximum selectivity toward CO oxidation that ap-
(i) from observation of zero-order oxygen pressure depen-
dence [41], and
(ii) by oxygen-exchange measurements between C O and
peared as a function of temperature in the case of Pt/
Al2O3 catalyst (Fig. 1) has been explained by competitive
Langmuir–Hinshelwood kinetics [12,16,28]: Pt adsorbs all
reactants and oxygen alternatively reacts with H2 or CO. The
selectivity toward CO oxidation would thus depend only on
1
6
1
8
O-predosed catalysts [39].
The results observed on Cl-containing samples also sup-
port the idea of a noncompetitive mechanism at the metal/
support interface. The presence of chloride ions in the vicin-
ity of Pt particles remarkably reduces the oxygen spillover
activity [39,47,48]. Accordingly, the CO-oxidation activity
decreased with increasing the Cl content, particularly at low
temperature and lower λ values, where oxygen mobility can
be the important factor (Fig. 3). However, oxygen conver-
sion remained the same, indicating that the oxygen excess
was used up for the hydrogen oxidation. Accordingly, the
occurrence of H2 oxidation does not need oxygen coming
from the support, because H is less strongly adsorbed than
O on platinum [51,52]. It is not the case for CO. In fact, θCO
is close to saturation at low temperature [28], where the ef-
fect of Cl was more pronounced. So, an oxygen flux coming
from the support can effectively promote the CO oxidation in
this temperature range. This phenomenon leads to two pos-
sible conclusions:
the ratio of θCO to θH . We recall that the heat of CO ad-
2
sorption on Pt is about 17 kJ/mol higher than the value for
hydrogen [49]. As a consequence the θCO:θH ratio decreases
2
with increasing temperature and the selectivity toward CO
oxidation as a function of temperature should continuously
decrease. The CO coverage is reported to be close to sat-
◦
uration in the low temperature range (90–180 C) [12,28].
Calculation upon DRIFT results showed that for a CO par-
tial pressure of 10 kPa, θCO is higher than 0.9θsaturation up
◦
to 200 C [28]. At the same time the catalytic activity in
the CO oxidation of Pt/Al2O3 is poor at low temperature
◦
(
T < 150 C) both in the absence [12] and in the presence
of H2 (Fig. 1). Hydrogen oxidation, in turn, is instantaneous
even at room temperature over Pt/Al2O3 [50]. Thus, at low
temperature, with poor catalytic activity, hydrogen oxida-
tion is preferred, even though θH is very small (compared
2
to θCO [16]). At this temperature, the rate constant is the
limiting factor and it is obviously higher for H2 [50]. In-
creasing the temperature leads to an increase in the total
(
i) CO oxidation preferentially occurs at the metal/support
interface, and
catalytic activity and since θCO:θH is high; the selectivity
(ii) hydrogen may be preferentially oxidized by separate
routes, on metal particles or probably on the support
alone.
2
toward CO oxidation increases (Fig. 1). On the contrary,
◦
at a higher temperature (T > 190 C) O2 conversion is to-
tal and accessible oxygen becomes rate limiting [12]. The
CO coverage is not any more close to saturation [16] and
In fact, ceria and CeO2–ZrO2 mixed oxides can adsorb hy-
drogen even in the absence of noble metal particles [53]. Hy-
drogen can then desorb both in the form of H2 (reversible ad-
sorption) and H2O (irreversible adsorption) [53]. This later
process means, indeed, hydrogen oxidation on the support.
so θCO:θH decreases remarkably. From this point hydrogen
2
and CO compete for oxygen. Their coverage ratio governs
the selectivity that should decrease.
In the case of ceria and ceria–zirconia-supportedsamples,
no maximum was found (Fig. 2). Since the active phase is
platinum in all cases, one might expect to use the same com-
petitive Langmuir–Hinshelwood kinetics to describe these
systems too. On the other hand, ceria and cerium–zirconium
mixed oxides are reducible supports, active in oxygen stor-
age [34–46], and a noncompetitive Langmuir–Hinshelwood
mechanism can be imagined on the metal/oxide interface.
It means reaction between CO adsorbed on Pt particles
and oxygen activated on the support. Such a noncompet-
itive mechanism was proposed for the PROX reaction on
Fe-promoted Pt/Al2O3 [18], bimetallic PtSn catalysts [30],
and Au/Fe2O3 [23]. The heat of adsorption of the differ-
ent molecules on Pt particles is, indeed, hardly affected by
the presence of ceria support [51,52]. On the contrary, at the
metal–ceria interface the kinetics for oxygen spillover may
play a role that is more important than the heat of adsorp-
tion measured under equilibrium conditions. Accordingly,
the participation of surface oxygen from the support during
the low-temperature CO oxidation on ceria-supported noble
metal catalysts has already been proved both:
◦
Such reaction was observed above 130 C [53].
Platinum is claimed to be an active catalyst in the water–
◦
gas-shift reaction above 200 C. Pt-on-ceria showed partic-
ularly high activity [54]. The effect of water has been stud-
ied indirectly: increasing O2 excess results in an increasing
amount of water in the reactor because the excess oxygen
reacts with hydrogen (Figs. 1 and 2). This water, formed by
H2 oxidation, can react with CO in a WGS reaction. This
process would virtually result in higher selectivity toward
◦
CO2 formation. Accordingly, at λ = 2 and above 200 C,
we observed that the selectivity toward CO oxidation on the
Pt/CeO2 sample is about constant or slightly increasing with
temperature, while it decreases at lower λ values (Fig. 2).
In summary, all processes are possible, i.e.,
(
i) competitive Langmuir–Hinshelwood CO and hydro-
gen oxidation on Pt particles, mostly in the case of
Pt/Al O ,
2
3
(ii) noncompetitive Langmuir–Hinshelwood mechanism
on the metal/oxide interface, in the case of reducible
support,

A. Wootsch et al. / Journal of Catalysis 225 (2004) 259–266
265
(
iii) hydrogen direct oxidation on the ceria(–zirconia) sup-
port, and
(
iv) water–gas-shift reaction at high temperature, particu-
larly on ceria-containing samples.
The second reaction route is predominant on Pt-on-ceria–
◦
zirconia samples at low temperatures (90–130 C) and CO
is preferentially oxidized rather than hydrogen. This process
was selectively blocked by Cl ions. At higher temperatures,
H2 oxidation on the ceria support may also be an important
process. The coexistence of these mechanisms
(a)
(
i) can be the reason for the high activity of ceria-supported
catalysts, and
(
ii) can explain the continuous increase in the selectivity
with decreasing temperature.
4
.2. Optimum catalytic formulation
One of the critical questions is: “What is the best PROX
catalyst?” Of course an ideal PROX catalyst
(
i) would be 100% selective toward CO oxidation when
λ ꢀ 1, and
(
ii) would convert all CO when λ ꢁ 1.
(b)
We suggest that an effective comparison between the differ-
ent catalysts can easily be done in a visualization such as
Fig. 4. The ideal PROX catalyst would result in a vertical
line at x = 0 (y axis), meaning 0% hydrogen loss, and then
in a horizontal line at 100% CO conversion. This later case
is possible only when λ ꢁ 1.
Fig. 4. Conversion of carbon monoxide as a function of the hydrogen loss
conversion) on Pt/Al O (a) and on the Cl-free Pt/CeO catalyst (b).
Data in (a) correspond to Fig. 1 and in (b) to Fig. 2. The reaction temperature
increased in the direction of the thin arrows. Results above the thick line,
“prohibited region,” cannot be reached under any reaction conditions. Thick
arrows: optimal operating point in a multistage reactor system; see text.
(H
2
2
3
2
Fig. 4 shows the conversion of CO as a function of the
hydrogen loss for Pt/Al2O3 (Fig. 4a) and for the Cl-free
Pt/CeO2 (Fig. 4b). Data obtained at the same oxygen excess
are connected forming a line at “iso-lambda.” The highest
CO conversion values at a given H2 loss are linked by a thick
line (Fig. 4). Points above this line cannot be reached under
the selected reaction conditions; so we called this area the
In fact, the maximum CO conversion (selectivity) is rather
contact time independent in a wide range [12,16]. The condi-
tions we used (100 mL/min total flow, 100 mg catalyst) were
in this range, corresponding well to that commonly reported
in the literature [12,14,16,22,25,28,29,33]. Furthermore, us-
ing more catalyst would not lead to higher CO conversion
since the oxygen conversion is total at maximum CO conver-
sion (Figs. 1 and 2). The selectivity increased monotonously
with decreasing temperature on Pt/ceria (Fig. 2). However,
the CO conversion cannot be further increased by decreas-
ing the temperature because of the sharp light off observed
in this case (Fig. 2). As a conclusion, a visualization such
as Fig. 4 is characteristic for a PROX catalyst at a given CO
inlet concentration (5% in our case).
On Pt/Al2O3 a maximum in selectivity is observed as a
function of temperature (Fig. 1). Such maxima are also visi-
ble in Fig. 4a. On the other hand, in the case of Pt/CeO2 cata-
lyst, straight iso-lambda lines were formed (Fig. 4b). In both
cases any selected CO conversion can be reached at various
λ values and different temperatures. However, it is obvious
that one should work close to the thick line, at low hydrogen
loss. Nevertheless, this line converges to 100% CO conver-
sion but does not reach it under the examined experimental
“
prohibited region.” We believe that the iso-lambda curves,
observed under the same conditions on different catalysts,
can serve for comparison purposes. The prohibited region
was somehow smaller in the case of ceria-supported Pt cata-
lyst than on Pt/Al2O3 (Fig. 4).
It must be noted that the extent of the prohibited region
also depends on the reaction conditions, and not only on the
catalyst nature. We used 5% CO in the inlet stream. Obvi-
ously, lower inlet CO concentration would result in a curve
being closer to the y axis, since (at a constant λ value) lower
amounts of O2 would be added to the reaction mixture, re-
sulting in lower loss in H2. On the other hand, the absolute
value of CO determines the surface coverage of the cata-
lyst [12,28]. If the CO concentration decreased to a value
when θCO is not close to saturation any more, selectivity to-
ward CO oxidation would decrease. Other operating parame-
ters, in turn, hardly affect the area of the prohibited region.

266
A. Wootsch et al. / Journal of Catalysis 225 (2004) 259–266
conditions. By extrapolating the line one may estimate that
a minimum 10% hydrogen would be lost at 99% CO conver-
sion. This problem can be solved by applying a multistage
reactor adding extra oxygen in between the two stages [14].
For example, two subsequent PROX reactors, both working
at 90% CO conversion, would result in about 99% CO con-
version at a much lower hydrogen loss. For minimizing the
hydrogen loss, one must use consecutive reactors operated
at a point just before the H2 loss increases sharply while CO
conversion levels up (thick arrows in Fig. 4). In the case of
ceria-supported Pt this point is around ∼ 80% CO conver-
sion at λ ≈ 1, while in the case of Pt on Al2O3 it is around
[7] F. Auprêtre, C. Descorme, D. Duprez, Catal. Commun. 3 (2002) 263.
[
8] C.D. Dudfield, R. Chen, P.L. Adcock, Int. J. Hydrogen Energy 26
2001) 763.
9] S.H. Lee, J. Han, K.-Y. Lee, J. Power Sources 109 (2002) 394.
(
[
[
[
10] S.H. Oh, R.M. Sinkevitch, J. Catal. 142 (1993) 254.
11] Y.-F. Han, M.J. Kahlich, M. Kinne, R.J. Behm, PCCP 4 (2002) 389.
[12] M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Catal. 171 (1997) 93.
[13] O. Korotkikh, R. Farrauto, Catal. Today 62 (2002) 249.
14] H. Igarashi, H. Uchida, M. Suzuki, Y. Sasaki, M. Watanabe, Appl.
[
Catal. 159 (1997) 159.
[
[
[
15] M.M. Schubert, A. Gasteiger, R.J. Behm, J. Catal. 172 (1997) 256.
16] H.D. Kim, M.S. Lim, Appl. Catal. A 224 (2002) 27.
17] A. Manaslip, E. Gulari, Appl. Catal. B 37 (2002) 17.
[18] X. Liu, O. Korotkikh, R. Farrauto, Appl. Catal. A 226 (2002) 293.
19] I.H. Son, M. Shamsuzzoha, A.M. Lane, J. Catal. 210 (2002) 460.
[
[
[
[
[
6
0% CO conversion at λ ≈ 0.8. Fig. 4, thus, clearly shows
the optimal operating point at a given inlet CO concentration
thick arrows). The CO conversion is not sufficient here, but
20] I.H. Son, A.M. Lane, Catal. Lett. 76 (2001) 151.
21] S. Özkara, A.E. Aksoylu, Appl. Catal. A 251 (2003) 75.
22] G.K. Bethke, H.H. Kung, Appl. Catal. 194–195 (2000) 43.
23] M.J. Kahlich, H. Gasteiger, R.J. Behm, J. Catal. 182 (1999) 430.
[24] R.J.H. Grisel, B.E. Nieuwenhuys, J. Catal. 199 (2001) 48.
[25] M.M. Schubert, V. Plzak, J. Garche, R.J. Behm, Catal. Lett. 76 (2001)
(
it can be improved in a second reactor stage. The efficiency
of the overall process, in turn, is the highest at this point.
Ceria can be an adequate support for Pt when one has to
lower the CO concentration in a hydrogen-rich stream by
using the PROX reaction at low temperature and/or with
a minimum hydrogen loss. Unfortunately, ceria-supported
platinum was found to be too active in hydrogen oxidation
as well, and cannot be considered as an effective PROX cat-
143.
[
[
26] B. Rohland, V. Plzak, J. Power Sources 84 (1999) 183.
27] H.-S. Oh, J.H. Yang, C.K. Costello, Y.M. Wang, S.R. Bare, H.H. Kung,
M.C. Kung, J. Catal. 210 (2002) 375.
[
[
[
28] M.M. Schubert, M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Power
Sources 84 (1999) 175.
29] G. Avgouropoulos, T. Ioannides, Ch. Papadopoulou, J. Batista, S. Ho-
cevar, H.K. Matralis, Catal. Today 75 (2002) 157.
30] M.M. Schubert, M.J. Kahlich, G. Feldmeyer, M. Hüttner, H.A. Hack-
enberg, H.A. Gasteiger, R.J. Behm, PCCP 3 (2001) 1123.
31] Y.-F. Han, M.J. Kahlich, M. Kinne, R.J. Behm, PCCP 4 (2002) 389.
32] A. Wörner, C. Friedrich, R. Tamme, Appl. Catal. A 245 (2003) 1.
33] Y. Hasegawa, A. Ueda, K. Kusakabe, S. Morooka, Appl. Catal. A 225
◦
alyst above 130 C. Nevertheless, ceria can be considered as
a beneficial additive to supported Pt catalysts [21]. Pt/CeO2-
(
–ZrO2) catalysts could also be introduced as an effective
part in a multistage PROX reactor system.
[
[
[
(
2002) 109.
Acknowledgments
[
34] Y. Madier, C. Descorme, A.M. Le Govic, D. Duprez, J. Phys. Chem.
B 103 (1999) 10999.
Most of the present work was carried out during the
postdoctoral fellowship of Attila Wootsch in Poitiers, spon-
sored by both the French Ministry of Foreign Affairs (from
[
35] A. Trovarelli, Catal. Rev.-Sci. Eng. 38 (1996) 439.
[36] A. Galdikas, C. Descorme, D. Duprez, Solid State Ionics 166 (2004)
147.
[
[
[
[
[
37] Y.Y. Yung-Fang, J. Catal. 87 (1984) 152.
1
October 2001 to 31 December 2001) and by the Ré-
38] S. Johansson, L. Österlund, B. Kasemo, J. Catal. 201 (2001) 275.
39] R. Taha, D. Martin, S. Kacimi, D. Duprez, Catal. Today 29 (1996) 89.
40] T. Bunluesin, H. Cordatos, R.J. Gorte, J. Catal. 157 (1995) 222.
41] E. Bekyarova, P. Fornasiero, J. Ka sˆ par, M. Graziani, Catal. Today 45
(1998) 179.
gion Poitou-Charentes (from 15 January 2002 to 15 No-
vember 2002). Additional mutual visits were part of the
cooperation between the Hungarian Academy of Sciences
and the CNRS. These financial supports are gratefully ac-
knowledged. Thanks to Dr. Sumeya Bedrane, Dr. Christelle
Carnevillier, and Ms. Mélanie Pajon for their help in catalyst
preparation.
[42] S. Bedrane, C. Descorme, D. Duprez, Catal. Today 73 (2002) 233.
[43] S. Bedrane, C. Descorme, D. Duprez, Catal. Today 75 (2002) 401.
[44] C. Descorme, Y. Madier, D. Duprez, J. Catal. 196 (2000) 167.
[45] A. Holmgren, B. Andersson, J. Catal. 178 (1998) 14.
[46] D. Martin, D. Duprez, J. Phys. Chem. 101 (1997) 4428.
[47] D.I. Kondarides, X.E. Verykios, J. Catal. 174 (1998) 52.
References
[48] S. Bernal, J.J. Calvino, G.A. Cifredo, J.M. Rodríguez-Izquierdo, V.
Perrichon, A. Laachir, J. Catal. 137 (1992) 1.
[
49] B. Sen, M.A. Vannice, J. Catal. 130 (1991) 9.
[
1] A.J. Appleby, F.R. Foulkes, Fuel Cell Handbook, Van Nostrand Rein-
hold, New York, 1989.
[50] M.T. Janicke, H. Kestenbaum, U. Hagendorf, F. Schuth, M. Fichtner,
K. Schubert, J. Catal. 191 (2000) 282.
[
2] W. Vielstich, T. Iwasita, in: G. Ertl, H. Knözinger, J. Weitkamp (Eds.),
Handbook of Heterogeneous Catalysis, vol. 4, Verlag Chemie, Wein-
heim, 1997, p. 2090.
[
[
51] G.S. Zafiris, R.J. Gorte, Surf. Sci. 276 (1992) 86.
52] T. Chafik, O. Dulaurent, J.L. Gass, D. Bianchi, J. Catal. 179 (1998)
503.
[
[
[
[
3] R.A. Lemons, J. Power Sources 29 (1990) 251.
4] J.M. Zalc, D.G. Löffler, J. Power Sources 111 (2002) 58.
5] J.N. Armor, Appl. Catal. 176 (1999) 159.
[53] S. Bernal, J.J. Calvino, G.A. Cifredo, J.M. Gatica, J.A.P. Omil, J.M.
Pintado, J. Chem. Soc., Faraday Trans. 89 (1993) 3499.
[54] J. Barbier Jr., D. Duprez, Appl. Catal. B 3 (1993) 461.
6] F. Joensen, J.R. Rostrup-Nielsen, J. Power Sources 105 (2002) 195.